DR-Transformer: A Multi-Features Fusion Framework for Tropical Cyclones Intensity Estimation

Round 1

Reviewer 1 Report

I have given this paper high marks, for what appears to be the successful implementation of their approach to calculating and predicting tropical storm intensity. 

A minor suggestion: to more fully explain the contents of tables and figures in each respective caption.

To make the paper more relevant to forecasters or potential users of the approach, it would be useful to:

1) Compare their results to dynamical forecasts.

2) To evaluate forecast changes rather than just absolute values. 

3) Another approach would be to calculate the Equitable Threat Score or Brier score for threshold exceedance values.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors have proposed a estimation and prediction method for tropical cyclone (TC) intensity.

Two types of features are used: Distance-consistency (DC) and Rotation-invariance (RI) features, extracted from cyclone images.

 

The proposed method can better learn the contours, structures, and other visual features of each TC image. DC features can reduce the estimation error between adjacent intensities, and RI features can eliminate feature deviation caused by conditions in which the images are obtained and TC rotation.

 

The authors experiments show that the final result, low error values are obtained, is better than any previously reported method trained on the tropical cyclone dataset.

 

A good contribution is that the proposed method is applied to the practical cyclone intensity estimation.

 

In their future work, the authors will extend the proposed feature extraction method for the most used numerical predictors in the meteorological field

 

I have some reviewer notes:

 

In the “Introduction” part, Lines 29-31. What are the error rates of the cited methods? It will be good to include values in your comments.

 

It will be good for data analytics, what is the data processing time using your and other methods. It will be good to comment it, for example in “Discussion” part.

 

The “Conclusion” part is very short. You can start the conclusion with your results and finish with the general problem of which your work is a part.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.